Post on 11-Mar-2021
Dynamical control of the mesosphere by orographic and non-orographic gravity wave drag during the extended northern winters of 2006 and 2009
Article
Published Version
McLandress, C., Scinocca, J. F., Shepherd, T. G., Reader, M. C. and Manney, G. L. (2013) Dynamical control of the mesosphere by orographic and non-orographic gravity wave drag during the extended northern winters of 2006 and 2009. Journal of the Atmospheric Sciences, 70. pp. 2152-2169. ISSN 1520-0469 doi: https://doi.org/10.1175/JAS-D-12-0297.1 Available at http://centaur.reading.ac.uk/33007/
It is advisable to refer to the publisher’s version if you intend to cite from the work. See Guidance on citing .Published version at: http://dx.doi.org/10.1175/JAS-D-12-0297.1
To link to this article DOI: http://dx.doi.org/10.1175/JAS-D-12-0297.1
Publisher: American Meteorological Society
Publisher statement: © Copyright [July 2013] American Meteorological Society (AMS).
All outputs in CentAUR are protected by Intellectual Property Rights law, including copyright law. Copyright and IPR is retained by the creators or other copyright holders. Terms and conditions for use of this material are defined in the End User Agreement .
www.reading.ac.uk/centaur
CentAUR
Central Archive at the University of Reading
Reading’s research outputs online
Dynamical Control of the Mesosphere by Orographic and Nonorographic GravityWave Drag during the Extended Northern Winters of 2006 and 2009
CHARLES MCLANDRESS
Department of Physics, University of Toronto, Toronto, Ontario, Canada
JOHN F. SCINOCCA
Canadian Centre for Climate Modelling and Analysis, Victoria, British Columbia, Canada
THEODORE G. SHEPHERD
Department of Meteorology, University of Reading, Reading, United Kingdom
M. CATHERINE READER
University of Victoria, Victoria, British Columbia, Canada
GLORIA L. MANNEY
NorthWest Research Associates, and New Mexico Institute of Mining and Technology, Socorro, New Mexico
(Manuscript received 23 October 2012, in final form 10 December 2012)
ABSTRACT
A version of the CanadianMiddle AtmosphereModel (CMAM) that is nudged toward reanalysis data up
to 1 hPa is used to examine the impacts of parameterized orographic and nonorographic gravity wave drag
(OGWD and NGWD) on the zonal-mean circulation of the mesosphere during the extended northern
winters of 2006 and 2009 when there were two large stratospheric sudden warmings. The simulations are
compared to Aura Microwave Limb Sounder (MLS) observations of mesospheric temperature and carbon
monoxide (CO) and derived zonal winds. The control simulation, which uses both OGWD and NGWD, is
shown to be in good agreement with MLS. The impacts of OGWD and NGWD are assessed using simu-
lations in which those sources of wave drag are removed. In the absence of OGWD the mesospheric zonal
winds in the months preceding the warmings are too strong, causing increased mesospheric NGWD, which
drives excessive downwelling, resulting in overly large lower-mesospheric values of CO prior to the
warming. NGWD is found to be most important following the warmings when the underlying westerlies are
too weak to allow much vertical propagation of the orographic gravity waves to the mesosphere. NGWD is
primarily responsible for driving the circulation that results in the descent of CO from the thermosphere
following the warmings. Zonal-mean mesospheric winds and temperatures in all simulations are shown to
be strongly constrained by (i.e., slaved to) the stratosphere. Finally, it is demonstrated that the responses to
OGWD and NGWD are nonadditive because of their dependence and influence on the background winds
and temperatures.
1. Introduction
The northern winters of 2006 and 2009 were punctu-
ated by two of the largest stratospheric suddenwarmings
on record. Fortuitously, this was also a period when
numerous research satellite instruments were in opera-
tion, providing an unprecedentedly detailed view of the
atmosphere from the upper troposphere to the lower
thermosphere (e.g., Manney et al. 2008a,b, 2009a,b;
Randall et al. 2006, 2009). From those observations the
following picture of the middle atmosphere emerged: In
themonths leading up to the warmings, which occurred in
late January, the stratosphere was near its climatological
Corresponding author address:CharlesMcLandress,Department
of Physics, University of Toronto, 60 St. George St., Toronto, ON
M5S 1A7, Canada.
E-mail: charles@atmosp.physics.utoronto.ca
2152 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
DOI: 10.1175/JAS-D-12-0297.1
� 2013 American Meteorological Society
state, with a stratopause and zonal-mean westerly jet
maximum located near 50 km. Following the warmings,
the middle and upper stratosphere became extremely
cold and undisturbed, and an elevated stratopause
formed in the upper mesosphere and slowly descended
to its climatological position by springtime. Chemical
species such as nitrogen oxides (NOx 5 NO 1 NO2)
and carbon monoxide (CO), which form in the upper
atmosphere and are transported downward in winter,
also underwent large changes, most notably strongly
enhanced descent following the warmings.
These mesospheric observations have led to a con-
certed effort on the part of modeling groups to simu-
late and understand the dynamics and transport during
and after these strong warming events. From the first
comparisons of the mesospheric observations to stan-
dard assimilated meteorological analyses, it became
apparent that models with lids higher than 80 km were
needed in order to simulate the reformation of the stra-
topause at high altitudes after the warmings (Manney
et al. 2008a).
In recent years, two high-lid data assimilation sys-
tems have been developed and used to study such warm-
ing events: the Canadian Middle Atmosphere Model
(CMAM; Polavarapu et al. 2005) and the Navy Oper-
ational Global Atmospheric Prediction-Advanced Level
Physics High Altitude (NOGAPS-ALPHA; Eckermann
et al. 2009). One of the most remarkable findings to
emerge was that the descent of the elevated stratopause
following the warmings could be realistically reproduced
without the assimilation of mesospheric temperature
observations provided that the effects of gravity wave
drag (GWD)were included, as was demonstrated byRen
et al. (2011) using the CMAM data assimilation system.
Siskind et al. (2010) used a different approach by per-
forming short-term forecasts initialized from an analysis
of the NOGAPS-ALPHA system, but came to a similar
conclusion. Both studies demonstrated that nonoro-
graphic GWD was responsible for the realistic meso-
spheric temperature response following the warmings, in
support of the earlier suggestion by Ren et al. (2008) that
the zonal-mean circulation in the mesosphere is slaved to
the lower atmosphere through GWD.
Two recent studies using free-running (i.e., con-
strained only by prescribed sea surface temperatures
and sea ice) versions of high-lid models have elucidated
the causes of several key dynamical features of warming
events like those in 2006 and 2009—in particular, the
elevated stratopause and its subsequent descent. Using
the Whole Atmosphere Community Climate Model
(WACCM), Limpasuvan et al. (2012) argued that the
formation of the elevated stratopause and its initial
descent were due to mesospheric planetary wave drag.
Hitchcock and Shepherd (2013) used a time-dependent
zonally averaged quasigeostrophic model driven by the
zonal-mean torques from a free-running version of
CMAM to diagnose the relative importance of the
different types of wave drag and the radiative forcing,
including their induced meridional circulations, which
are a significant component of the mesospheric response
to the sudden warming. This methodology allowed a
complete quantitative attribution of the response to the
different forcings. Their analysis showed that the meso-
spheric cooling immediately after the warming was due
to a combination of radiative cooling and a transient
circulation induced by the resolved wave drag in the
stratosphere. As in the Limpasuvan et al. (2012) study,
Hitchcock and Shepherd found that the formation of the
elevated stratopause was initiated by the drag exerted by
planetary waves propagating up from below. Both studies
reconfirmed the findings of Siskind et al. (2010) and Ren
et al. (2011) that the descent of the elevated stratopause
during the recovery period was primarily driven by non-
orographic GWD.
While free-running models are important for under-
standing the mechanisms responsible for the observed
mesospheric response to warmings such as these, their
model states are not sufficiently constrained to repro-
duce specific realizations of the atmospheric flow and
thus cannot be compared quantitatively to observations
of a particular event on a day-by-day basis. Since the
winds and temperatures in the lower atmosphere have
a strong influence on the propagation and absorption of
gravity waves, free-running models likewise cannot be
expected to reproduce the actual gravity wave fluxes
entering the mesosphere for a particular event. It is only
by constraining the winds and temperatures in the lower
atmosphere (e.g., by using data assimilation) that the
parameterized (and resolved) gravity wave fluxes could
match the true fluxes. Doing so would then enable
a quantitative evaluation of themesospheric GWD in the
model, which would be achieved by comparing the sim-
ulated large-scale winds and temperatures in the meso-
sphere to observations. Such a study has yet to be done.
Moreover, none of the previous modeling studies exam-
ined the time period when the vortex was developing,
focusing instead on the warmings and their aftermath,
when the effects of nonorographic GWDwere dominant.
Thus, the extent to which orographic GWD controls the
mesospheric circulation is unclear. It is also unknown
whether the mesosphere is slaved to the lower atmo-
sphere in late fall/early winter and, if it is, whether it is
realized through orographic GWD, planetary wave drag,
or a combination of the two.
Herein, we describe results from a study designed to
investigate these issues using a version of CMAM in
JULY 2013 MCLANDRES S ET AL . 2153
which a simple relaxation procedure is employed to
constrain the model to follow reanalysis data up to
1 hPa. Since the model includes interactive chemistry
and online transport, we can examine the descent of
chemical constituents like CO and NOx. By validating
our simulations against mesospheric observations of
temperature and CO, as well as derived zonal winds,
from the Aura Microwave Limb Sounder (MLS) we
can quantify the role of parameterized GWD in driving
the mesospheric circulation in the model and by in-
ference in the real atmosphere. We assess the relative
roles of orographic and nonorographic GWD using a
set of sensitivity experiments in which the two types of
drag are turned off, separately and together. This also
enables us to assess the additivity of the responses to
orographic and nonorographic GWD, which has hitherto
not been done.
The advantage of our nudging approach is that the
lower atmosphere (i.e., the portion below 1 hPa that is
nudged) is to a large extent unaffected by the removal of
one or another type of GWD, thus keeping the resolved
waves and the momentum fluxes from the other GWD
scheme largely unchanged throughout the lower atmo-
sphere. This permits the separation of cause and effect in
the mesosphere, which is not possible when the lower
atmosphere is unconstrained, as in the case of the GWD
sensitivity forecasts of Siskind et al. (2010). (Forecasts
can only be run for short times before the stratospheric
flow has departed so far from the observed state that the
gravity wave fluxes entering the mesosphere are signif-
icantly changed.) Although the CMAM data assimila-
tion system could be used here, the cost of performing
the integrations would be prohibitive and there turns out
to be little or no advantage for this type of study com-
pared to our simple relaxational approach.
The paper is organized as follows. In section 2 we
describe the data used. This includes a very brief dis-
cussion of theMLS data and amore in-depth discussion
of CMAM and our simulations. We include here a
fairly detailed description of the GWD parameteriza-
tions since they are crucial components of our study. In
section 3 we present the results, starting with a de-
scription of the MLS observations and followed by a
discussion of the control simulation for the ‘‘extended’’
winters (i.e., September–April) of 2006 and 2009. We
then discuss the GWD sensitivity experiments, which
enable us to understand the relative roles of orographic
and nonorographic GWD in determining the zonal-
mean dynamics and transport in the mesosphere. We
end this section with a discussion of the additivity of the
responses to orographic and nonorographic GWD. In
section 4 we summarize our main results and discuss
some implications of our findings.
2. Data
a. MLS observations
The MLS instrument on the Aura satellite has pro-
vided a nearly continuous set of measurements of tem-
perature and trace gases in the middle atmosphere from
August 2004 through the present. Here we use version
3.3 temperature and CO data (Livesey et al. 2011), as
well as zonal winds derived from the MLS geopotential
heights using a balanced wind formulation (Randel
1987) as described by Manney et al. (2008a). Daily av-
erage zonal means on a 28 latitude grid are used.
b. CMAM simulations
CMAM is a chemistry–climate model extending from
the earth’s surface to about 100 km. Detailed descrip-
tions of the model and stratospheric chemistry scheme
are given, respectively, in Scinocca et al. (2008) and de
Grandpr�e et al. (2000) [with updates to the chemistry
provided in Jonsson et al. (2004)]. The version of the
model we use has a triangular spectral truncation of T47,
corresponding to a 3.758 horizontal grid on which the
physical parameterizations are evaluated. There are 71
levels in the vertical, with a resolution varying from
several tens of meters in the lower troposphere to about
2.5 km in the mesosphere. To prevent wave reflections
at the model lid a Rayleigh friction sponge is applied
above about 85 km to the deviations of the horizontal
winds from the zonal mean. Monthly and annually vary-
ing sea surface temperatures and sea ice distributions are
prescribed using observations. The radiative forcings and
chemical boundary conditions are the same as those used
in SPARC CCMVal (2010).
Orographic GWD (OGWD) and nonorographic
GWD (NGWD) are parameterized using the schemes
of Scinocca andMcFarlane (2000) and Scinocca (2003),
respectively. The OGWD scheme employs two vertically
propagating zero-phase-speed gravity waves (GWs) to
transport the horizontal momentum contained in all
waves directed into the half-space to the left and right of
the large-scale (i.e., resolved) horizontal velocity vector
at the launch layer, which extends from the surface to
the height of the subgrid topography. The orientation
and magnitude of the momentum flux carried by these
two waves is a function of the near-surface wind speed,
its direction relative to the orientation of the subgrid
topography, and the static stability in the launch layer.
The two relevant tunable parameters in the OGWD
scheme are the multiplicative-scale factor, G(y) 5 0.65,
and the inverse critical Froude number, Frcrit5 0.375. In
general, G(y) scales the total vertical flux of horizontal
momentum and Frcrit determines the breaking height.
However, Frcrit also has influence over the total amount
2154 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
of launch momentum flux because it is used to set the
maximum amplitude of the parameterized waves at
launch (i.e., waves are not permitted to exceed their
breaking amplitude at launch). The above values of
G(y) and Frcrit have been tuned for polar-ozone chem-
istry studies in CMAM since they produce reasonable
zonal-mean zonal winds and polar temperatures in the
winter lower stratosphere (Scinocca et al. 2008).
The NGWD scheme employs a spectrum of nonzero-
phase-speed GWs propagating horizontally in the four
cardinal directions. The launch level vertical wave-
number (m) energy spectral density is prescribed to be
the so-called Desaubies spectrum, being independent of
m form�m* and with anm23 dependence form�m*,
where m* 5 2p/(1 km), and is identical for all azimuths
and spatially uniform. The total nonorographic GW
Eliassen–Palm flux in each azimuth at the launch level
(;125 hPa) is the primary tuning parameter and is
typically set to 4.243 3 1024 Pa (as it is here) since it
produces a reasonable seasonal evolution of the zonal-
mean zonal winds and temperatures in the mesosphere
in CMAM. Since the launch spectrum is isotropic, the
net zonal and meridional Eliassen–Palm fluxes at the
launch level are zero.
As the parameterized orographic and nonorographic
GWs propagate upward they are subject to both critical
level filtering and nonlinear saturation. The latter is
treated in the NGWD scheme by not permitting the en-
ergy spectrum to exceed the observedm23 form, while in
the OGWD scheme a convective instability threshold is
applied. In the uppermost model layer all remaining GW
momentum flux is deposited in order to prevent any
spurious downward influence (Shepherd and Shaw 2004).
The parameter settings used for both GWD schemes are
identical to those used in the CMAM contribution to
SPARC CCMVal (2010).
Below 1 hPa the horizontal winds and temperatures
are nudged (i.e., relaxed) to the 6-hourly horizontal
winds and temperatures from the European Centre for
Medium-Range Weather Forecasts (ECMWF) Interim
Re-Analysis (ERA-Interim; Dee et al. 2011). A key
aspect of the relaxational approach adopted here is that
it is designed to primarily constrain the synoptic space
and time scales in the model, which are well represented
in the reanalysis data. This is accomplished by performing
the relaxation in spectral space and applying it only to
horizontal scales with n# nmax 5 21, where n is the total
wavenumber. The nudging tendency, which is applied to
the vorticity, divergence, and thermodynamic equations,
has the form 2(X2XR)/t0, where t0 5 24 h is the re-
laxational time scale and X and XR are, respectively, the
model and reanalysis spectral vorticity, divergence, or
temperature coefficients. Linear interpolation is used to
compute the reanalysis variables at intermediate times
between adjacent 6-h intervals. The values of nmax and
t0 were chosen so that the RMS differences between the
nudged temperature fields and ERA-Interim are on aver-
age roughly similar to the RMS differences between pairs
of different reanalysis datasets [i.e., ERA-Interim, 40-yr
ECMWF Re-Analysis (ERA-40), National Centers for
Environmental Prediction reanalysis (NCEP), and Na-
tional Centers for Environmental Prediction—Department
of Energy Reanalysis 2 (NCEP2)], where the latter differ-
ences provide an estimate of the uncertainty in the re-
analysis products (Merryfield et al. 2013).
To demonstrate the impact of the nudging on the
model temperatures, we present here results from several
short simulations using three different nudging configu-
rations. In the first the zonal means are nudged up to
1 hPa using the constant t0 and the deviations from the
zonal means are nudged up to 1 hPa using a height-
dependent relaxational time scale that is equal to t0 below
10 hPa and whose inverse is tapered to zero between 10
and 1 hPa. The tapering is used to minimize the possible
distortion of vertically propagating (resolved) waves that
might occur just above 1 hPa where the relaxational time
scale jumps to infinity. We will refer to this as tapered
nudging. The two other configurations are oneswhere the
nudging is applied only up to 10 hPa and only up to
100 hPa, without any tapering.
The results of these three experiments are presented
in Fig. 1, which shows polar-cap average (708–908N)
temperatures in the stratosphere [10 hPa (32 km); top
panel] and mesosphere [0.046 hPa (70 km); bottom
panel] for the winter months of 2006. For each experi-
ment an ensemble of three simulations was constructed
using perturbed initial conditions. The colored curves
denote the three CMAM ensembles, while the black
dotted curves denote the MLS observations. At 10 hPa
the agreement between the ensemble using tapered
nudging (red) and MLS is very good, as expected since
this is well within the nudging region where MLS and
ERA-Interim are in good agreement (not shown). The
agreement with MLS at 10 hPa worsens as the nudging
height is lowered to 10 hPa (blue) and 100 hPa (green).
At 70 km the ensemble using tapered nudging is still in
good agreement with MLS, which is remarkable given
that this is a full 20 km above the top of the nudging
region. Moreover, the spread of the three ensemble
members is very small, indicating that the polar-cap
temperatures in the mesosphere are tightly constrained
by (i.e., slaved to) the state of the lower atmosphere. As
the nudging height is lowered to 10 and 100 hPa, the
agreement with MLS at 70 km worsens and the inter-
ensemble spread increases, with the 100-hPa nudging
height yielding the poorest results.
JULY 2013 MCLANDRES S ET AL . 2155
The above analysis motivates the use of the tapered
nudging configuration in the following four ensembles of
simulations that form the backbone of this paper:
1) control, 2) without OGWD, 3) without NGWD, and
4) without GWD (i.e., with both OGWD and NGWD
turned off). The latter three are otherwise identical to the
control ensemble. The simulations extend from January
2004 to December 2010, but we present results for only
the extended winters of 2006 and 2009. Each ensemble
consists of three members, with the different members
generated by perturbing the initial conditions in 2004.
Unless stated otherwise, ensemble averages are shown.
All plotted results are daily means, computed from in-
stantaneous 6-hourly data. Log-pressure height (com-
puted using a constant scale height of 7 km) is used for
the vertical axes in all figures, as well as for specific alti-
tudes mentioned in the text.
3. Results
a. MLS observations
The black dotted curves in Fig. 2 show MLS polar-
cap temperatures at 70 km for the extended winters of
2006 and 2009. (The colored curves will be discussed in
due course.) The two winters exhibit similar temporal
behavior—namely, a moderate degree of day-to-day
variability from mid-October until near the time of the
major warmings (21 January 2006 and 24 January 2009),
cooling during and immediately following the warmings
(more pronounced in 2009), and a gradual and smooth
warming and then cooling from February to April. The
occurrence of a minor warming on 1 January 2006 ap-
pears to have resulted in the much less abrupt tempera-
ture drop at the time of the 2006 major warming.
Figures 3b, 3d, 4b, and 4d show time versus height
sections of MLS polar-cap temperatures and the corre-
sponding midlatitude average (408–808N) zonal winds for
the two extended winters. As in Fig. 2, the results are
qualitatively similar in the two years: Prior to the warm-
ings, the stratopause is located near its climatological
position at about 50 km, and the winds and temperatures
exhibit the typical day-to-day variability of the late-fall/
early-winter stratosphere. Immediately following the
warmings, the stratosphere is cold and undisturbed, and
a new stratopause forms in the upper mesosphere and
steadily descends to its climatological position by
springtime. The minor warming in early January 2006 is
seen as the first patch of easterlies in Fig. 3d.
Figures 3f and 4f show polar-cap CO volume mixing
ratios from MLS. The downward tilt of the contours in
fall/early winter and late winter is indicative of the de-
scent of CO from the thermosphere where it is produced
by the photolysis of CO2. Near or at the time of themajor
warmings, the signature of polar descent disappears
because of the strong horizontal mixing that occurs as
the vortex disappears, bringing in low values of CO from
middle latitudes. Following the warmings, CO descends
again until the final breakup of the polar vortex in
springtime. An important point to note here is that the
late-winter maximumof theMLSCOabove about 60 km
is substantially larger than the early-winter maximum
(see also Fig. 8, black curves). The main difference in the
temporal behavior of CO between the 2 years is that the
signature of early-winter descent in 2009 is apparent right
up until the major warming, while in 2006 it stops in early
January at the time of the minor warming, and undergoes
rapid day-to-day variations until the major warming.
While the synoptic structure of the twowarming events
was very different—the 2006 warming being character-
ized as a ‘‘vortex displacement’’ and the 2009 warming as
a ‘‘vortex split’’—the mesospheric response following
these warmings is seen to be very similar. This is consis-
tent with the finding of Hitchcock et al. (2013) that the
morphology and time scale of the extended recovery of
the stratosphere, which drives the mesospheric response,
is insensitive towhetherwarmings occur as displacements
FIG. 1. Polar-cap average temperatures at (a) 10 and (b) 0.046 hPa
for the winter of 2006. Colored curves denote the following three
ensembles of simulations (with each ensemble consisting of three
members): tapered nudging (red), nudging only up to 10 hPa (blue),
and nudging only up to 100 hPa (green). The black dots denote the
MLS observations. Days range from 1 Dec 2005 to 28 Feb 2006.
2156 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
or splits in the long-term observational record. It is also
consistent with the fact that in the free-running CMAM,
the mesospheric response to extended recoveries is the
same for displacement and split events (Hitchcock and
Shepherd 2013).
b. Control simulation
The red curves in Fig. 2 show the polar-cap tempera-
tures at 70 km for the three members of the control sim-
ulation ensemble. Aswith Fig. 1, the agreement withMLS
(black) is remarkably good, with differences less than
about 10 K. Figures 3a, 3c, 4a, and 4c show the corre-
sponding winds and temperatures for the ensemble av-
erage of the control simulation as a function of height and
time. Good agreement with the MLS observations is seen
throughout the entire height region above about 50 km
where the model is not nudged. This supports the main
conclusion of Ren et al. (2011) that it is not necessary to
assimilatemesospheric temperature data to reproduce the
observed mesospheric temperature response to the 2006
major SSW and the subsequent descent of the elevated
stratopause. However, our results extend that conclusion
to include the fall and early-winter months, which are not
discussed by Ren et al. (2011). The good agreement be-
tween the control simulation and MLS is more clearly
seen in Figs. 5a, 5b, 6a, and 6b, which show the corre-
sponding differences withMLS for the extended winters
of 2006 (left panels) and 2009 (right panels). Below
about 85 km the absolute differences are less than about
5 K and 5 m s21. In the region above about 85 km the
agreement is worse, but this is in the model sponge re-
gion where good agreement cannot be expected.
RMS differences between the control simulation and
the MLS polar-cap temperatures and midlatitude zonal
winds for the two extended winters combined are shown
in Fig. 7 (red). As in Figs. 5 and 6, the differences are less
than about 5 K and 5 m s21 over most of the plotted
domain. In addition, the three members of the control
ensemble exhibit very little spread, indicating that the
polar-cap temperatures and midlatitude average zonal
winds are slaved to the state of the lower atmosphere
over the full depth of the mesosphere for the entire ex-
tended winter periods. This lack of spread in the control
simulation ensemble is also seen in Fig. 2, with the ex-
ception of the time period during the major warming in
late January 2009 when the members differ from each
other by up to about 10 K.
The polar-cap CO from the control simulation is also
in good agreement with that from MLS (Figs. 3e,f and
4e,f). This is seen more clearly in Fig. 8, which shows
results at 60 km for the two years. The control simulation
CO (red) agrees to within about 10%–20% of the MLS
observations (black) and exhibits the same seasonal
variation, namely the two-peaked structure, with the
smaller peak in early winter and the larger peak about
40 days after the major warmings. Qualitative agreement
with the MLS CO observations in the mesosphere was
found using an earlier version of CMAM (Jin et al. 2009).
FIG. 2. Polar-cap average temperatures at 70 km for the extended winters of (a) 2006 and
(b) 2009: control simulation (red) and the simulationswithout orographicGWD(blue), without
nonorographic GWD (green), and without GWD (yellow). The black dotted curves denote the
MLS observations. The three members of each CMAM ensemble are shown. Days range from
1 Sep 2005 (2008) to 30 Apr 2006 (2009).
JULY 2013 MCLANDRES S ET AL . 2157
However, since that was a free-running version of the
model it was not possible to make a quantitative com-
parison on a day-to-day basis as we have done here.
c. Relative roles of orographic and nonorographicGWD
In this section we examine the relative roles of OGWD
and NGWD in producing the realistic zonal-mean me-
sospheric circulation found in the control simulation.
Here we explain why the control simulation exhibits
the same seasonal variation of mesospheric CO as seen
in the MLS observations—namely, the double-peaked
structure—with the late-winter peak being approxima-
tely twice as large as the early-winter peak.
Figures 5c–h and 6c–h show the differences between
CMAM and MLS polar-cap temperatures and mid-
latitude average zonal winds for the three sensitivity
experiments. The overall differences far exceed those
of the control simulation (top row). Consider first the
simulation without OGWD (second row). The largest
differences occur in late fall/early winter before the
major warmings: the modeled lower mesosphere is up to
35 K too cold and the upper mesosphere up to 25 K too
warm. (Note that the reasonably good agreement be-
tween this experiment and MLS seen in Fig. 2 is for-
tuitous since 70 km happens to be near the node in the
temperature difference dipole seen in Figs. 5c and 5d.)
Correspondingly, the modeled zonal winds are too
FIG. 3. (a),(b) Polar-cap average temperatures, (c),(d) zonal winds averaged from 408 to 808N, and (e),(f) polar-cap
average CO volume mixing ratios for the extended winter of 2006: (left) control simulation and (right) MLS ob-
servations. Contour intervals are 10 K and 10 m s21 in (a),(b) and in (c),(d), respectively; CO has units of ppmv and
a logarithmic contour interval. Temperatures less than 220 K and easterlies are shaded blue. Days range from 1 Sep
2005 to 30 Apr 2006. Missing MLS data not contoured. Above about 85 km there are few MLS CO data.
2158 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
strong, with differences approaching about 50 m s21 in
the mesosphere. In 2009 the large wind and tempera-
ture differences extend right up to the start of themajor
warming, which is in contrast to 2006when the differences
start to weaken about the time of the minor warming in
early January. In addition, the wind and temperature
differences are about a factor of 2 smaller after the major
warmings than before.
In the simulation without NGWD (Figs. 5e,f and 6e,f),
the differences are also large but, in contrast to the
simulation without OGWD, tend to be largest after the
major warmings, not before. The removal of NGWD
delays the descent of the stratopause (revealed by the
MLS temperatures in Figs. 3b and 4b) until later in the
winter when the lower-stratospheric westerlies have
increased sufficiently to allow the upward propagation
of the (parameterized) orographic GWs. There is
also a significant impact of NGWD at the beginning
and end of the extended winters when the circula-
tion is closer to summertime conditions. This is more
clearly seen in early September when the removal of
NGWD has resulted in negative zonal wind differ-
ences (i.e., easterlies) in the upper mesosphere (Figs.
6e and 6f).
In the simulation without GWD (Figs. 5g,h and 6g,h)
the differences with respect to MLS are largest, with
temperatures up to about 65 K too low and winds up to
about 85 m s21 too high. An elevated stratopause does
not form, and the stratopause reformation at its clima-
tological position near 50 km is delayed until March
when the effects of solar heating start to become im-
portant. Although the wind and temperature difference
fields exhibit the overall combined characteristics of the
simulations without OGWD and without NGWD, it is
FIG. 4. As in Fig. 3, but for 2009, with days ranging from 1 Sep 2008 to 30 Apr 2009.
JULY 2013 MCLANDRES S ET AL . 2159
clear that at certain times and places (e.g., early winter
between 50 and 70 km) the sum of the differences in the
second and third rows of Figs. 5 and 6 do not equal those
in the bottom row. This indicates that the responses to
OGWD and NGWD are not additive. This issue is in-
vestigated in section 3d.
The RMS wind and temperature differences (with re-
spect to MLS) for the three sensitivity experiments are
FIG. 5. Differences between MLS and CMAM polar-cap average temperatures for the extended winters of (left)
2006 and (right) 2009: (a),(b) control simulation and the simulations (c),(d) without orographicGWD, (e),(f) without
nonorographic GWD, and (g),(h) without GWD. Contour interval is 10 K, with blue (red) shading for negative
(positive) differences exceeding 5 K in magnitude. Data have been smoothed using a 3-day boxcar filter. Note that
the height range differs from that in Figs. 3 and 4.
2160 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
shown in Fig. 7. Theworst agreement is for the simulation
withoutGWD (yellow), where the RMS differences peak
at about 35 K and 50 m s21. As with the control ensem-
ble, the spread between the different members of the
sensitivity experiment ensembles is also very small. This
is also seen in Fig. 2, which shows polar-cap temperatures
at 70 km for the two extended winters. Note that time
series at other mesospheric heights (not shown) exhibit
the same lack of spread. The lack of spread in the
members of the ensemble withoutGWD is of particular
interest because it demonstrates that the slaving of the
mesosphere to the stratosphere is not only throughGWD.
FIG. 6. As in Fig. 5, but for zonal winds averaged from 408 to 808N. Contour interval is 10 m s21, with blue (red)
shading for negative (positive) differences exceeding 5 m s21 in magnitude.
JULY 2013 MCLANDRES S ET AL . 2161
In this case it is the resolved (planetary) wave drag that
is constraining the mesospheric temperatures before
the warmings. In the extended period after the warmings,
planetary wave activity in the stratosphere is very weak,
which is reflected in the simulations by extremely weak
resolved wave drag below about 80 km (not shown).
Thus, in the absence of GWD the mesospheric tem-
peratures quickly relax toward radiative equilibrium,
which explains the similarity of the three members of
that ensemble in the period following the warmings.
Figures 2 and 7 also demonstrate another important
point—namely, that slaving to the stratosphere does
not imply that the mesospheric circulation is realistic,
only that its temporal variation is constrained. The re-
alism of the mesospheric circulation depends on the re-
alism of the slaving mechanisms.
Returning to the time series of polar-cap CO (Fig. 8),
large differences are seen between the three sensitivity
experiments andMLS. In the simulation without OGWD
(blue) the early-winter peak is too strong and the late-
winter peak too prolonged. In the simulation without
NGWD(green) the agreement withMLS is better than in
the simulation without OGWD in early winter, but the
late-winter peak is too weak, too brief, and is shifted to
a later date, particularly in 2009. In the simulation with-
out GWD (yellow) CO is far too weak in October and
November and following the warmings.
In the next three figures we explain why the temporal
behavior of mesospheric CO is so different for the three
sensitivity experiments and why the agreement with
MLS CO is best for the control simulation. We start by
examining the residual vertical velocity w*, which is
largely controlling the concentration of CO over the
winter pole. Downward control (Haynes et al. 1991) is
then used to attribute the differences in w* to the dif-
ferent types of wave drag.
Figure 9 shows polar-capw* for the control simulation
and the three sensitivity experiments. In the upper me-
sosphere (i.e., above about 70 km) the control simulation
FIG. 7. RMS differences between MLS and CMAM for the
combined extended winters of 2006 and 2009 (1 Sep–30 Apr) for
the control simulation (red), and the simulations without oro-
graphic GWD (blue), without nonorographic GWD (green), and
without GWD (yellow): (a) polar-cap average temperatures and
(b) zonal winds averaged from 408 to 808N. All three members of
the four different ensembles are shown.
FIG. 8. Polar-cap average CO volume mixing ratios at 60 km for
the extended winters of (a) 2006 and (b) 2009: control simulation
(red) and the simulations without orographic GWD (blue), without
nonorographic GWD (green), and without GWD (yellow). Black
curves denote the MLS observations. Days range from 1 Oct 2005
(2008) to 30 Apr 2006 (2009).
2162 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
(Figs. 9a,b) exhibits weak downwelling in late fall/early
winter and strong downwelling in late winter following
the warmings, thus explaining the weak early-winter and
strong late-winter peaks of CO in Fig. 8. In the simulation
without OGWD (Figs. 9c,d) the late-fall/early-winter
downwelling in the upper mesosphere is a factor of 2
stronger than in the control, while in the region below it is
weaker. This is more clearly seen in Fig. 10a, which shows
FIG. 9. Polar-cap average residual vertical velocities for the extendedwinters of (left) 2006 and (right) 2009: (a),(b)
control simulation and the simulations (c),(d) without orographic GWD, (e),(f) without nonorographic GWD, and
(g),(h) without GWD. Contour interval is 3 mm s21, with blue (red) shading for downwelling (upwelling). Data have
been smoothed twice using a 5-day boxcar filter.
JULY 2013 MCLANDRES S ET AL . 2163
vertical profiles of w* for a 40-day period prior to the
2009 warming. It is the strong downwelling in the upper
mesosphere that is causing the unrealistically large
early-winter peak in CO in the simulation without
OGWD (Fig. 8). The anomalous upper-mesospheric
downwelling and lower-mesospheric upwelling that oc-
curs whenOGWD is switched off also results in adiabatic
warming and cooling, respectively, which explains the
vertical dipole structure in the temperature differences in
early winter (Figs. 5c,d). After the warmings the polar-
cap w* is similar in strength in the two simulations (Figs.
9a–d and 10c), but the tongue of downwelling is some-
what wider (in time) in the simulation without OGWD,
thus explaining why the late-winter peak of CO is wider
than in the control. In the simulation without NGWD
(Figs. 9e,f) the early-winter downwelling is similar to that
of the control, but the tongue of downwelling following
the warmings is narrower and in the 2009 winter does
not descend as far down, thus explaining the weak
late-winter peak of CO seen in Fig. 8. In the simulation
without GWD (Figs. 9g,h) the downwelling in the
lower mesosphere is very weak except in December
and January, whereas in the upper mesosphere there is
strong downwelling from October to mid-April. Since
there is no parameterized GWD in this simulation, w* is
largely being driven by resolved wave drag.
Figures 10b,d show the downward control (DC) es-
timates of the polar-cap w* (w) computed using the
different types of wave drag and averaged over the two
40-day periods. The good overall agreement between w
computed using the total wave drag (thick curves) andw*
(Figs. 10a,c) demonstrates the validity of the DC calcu-
lation, except in the model sponge. Note that the mag-
nitude of w is somewhat larger thanw*, as expected, since
w* does not have time to fully adjust to the steady-state
conditions assumed in the downward control. An exam-
ination of the contributions from the individual forcings
(thin curves) reveals that NGWD (solid) is largely re-
sponsible for the strong downwelling in the upper me-
sosphere prior to the warming (Fig. 10b) in the simulation
without OGWD (blue curve), with a secondary contri-
bution coming from resolved wave drag (dotted). The
latter is an indirect response to the removal of OGWD,
resulting from the increased westerlies above 50 km
(Fig. 6d) that allow planetary waves to propagate to
higher elevations where they dissipate, as inferred from
FIG. 10. Vertical profiles of polar-cap average residual vertical velocity averaged over 40-day periods (a) before
and (c) after the 2009 major SSW for the control simulation (red) and simulation without OGWD (blue). (b),(d)
Corresponding downward control estimates of w* computed using the total wave drag (thick solid), only nonoro-
graphic GWD (thin solid), only orographic GWD (dashed–dotted), and only resolved wave drag (dotted). The
averaging periods start 50 days before and 10 days after the date of the actual warming (24 Jan). Total wave drag is the
sum of the resolved wave drag and orographic and nonorographic GWD.
2164 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
large negative values of resolved wave drag above about
85 km in early winter (results not shown).1 The late-fall/
early-winter increase inNGWD in the upper mesosphere
in the simulation without OGWD is a direct consequence
of the increase in the strength of the mesospheric
westerlies seen in Fig. 6d, which has raised the region of
saturation of the westward-propagating nonorographic
GWs (not shown). This indicates that there is a strong
interplay between OGWD and NGWD in late fall/early
winter, mediated by their impact on the background
winds and temperatures.
In the 40-day period after the warming, downwelling is
driven primarily by NGWD (Fig. 10d), with OGWD be-
coming more important later on as westerlies strengthen
in the lower stratosphere (not shown). In comparison to
the period before thewarming, the downwelling driven by
OGWD after the warming is much weaker and occurs
higher up. Qualitatively similar results are found for the
winter of 2006.
d. Additivity of responses to orographic andnonorographic GWD
While the wind and temperature differences (with
respect to MLS) for the simulation without GWD
(Figs. 5g,h and 6g,h) appear to be qualitatively similar
to the sum of the differences of the simulations without
OGWD and without NGWD, there is clear evidence that
the responses to OGWD and NGWD are nonadditive.
To quantify this we introduce a diagnostic, which we shall
refer to as the GW additivity anomaly DX, defined by
DX5DXGWD2 (DXOGWD1DXNGWD), (1)
whereX represents a quantity like temperature or zonal
wind, DXGWD is the response of X to GWD (i.e., X for
the control minus X for the simulation without GWD),
DXOGWD is the response to OGWD (i.e., control minus
simulation without OGWD), and DXNGWD is the re-
sponse to NGWD (i.e., control minus simulation with-
out NGWD). If the response is additive, then DX 5 0.
Figure 11 shows the GW additivity anomalies for
midlatitude zonal wind DU (Fig. 11a) and polar-cap
temperature DT (Fig. 11b) for the winter of 2009. Qual-
itatively similar results are found for 2006. There are two
periods when the responses are nonadditive: the winter
months leading up to the major warming in late January,
and the month or so following the warming. In general,
DT is positive (peak values of about 25 K near 65 km)
andDU is negative (peak values of about245 m s21 near
85 km), with the sign difference and vertical offset of the
maxima being consistent with thermal wind balance.
The strength of the nonadditivity can be expressed as
the ratioRX5DXGWD/(DXOGWD1DXNGWD). For the
midlatitude zonal wind RU averaged from November
to January increases from about 1.2 at 65 km to about
1.8 at 85 km, indicating that the zonal wind response to
OGWD and NGWD is strongly nonadditive in the upper
mesosphere at this time. Additivity is only apparent
during the month following the major warming (as in-
dicated by the white area in Fig. 11). However, such ad-
ditivity is specious as it occurs during a period when both
the mesospheric OGWD and lower-mesospheric re-
solved wave drag are extremely weak.
Positive DT in Fig. 11b means that the temperature
response to GWD (DTGWD) exceeds that of the sum of
the separate responses (DTOGWD 1DTNGWD). To help
FIG. 11. GW additivity anomalies for (a) zonal winds averaged
from 408 to 808N (DU) and (b) polar-cap temperature (DT ) for the
winter of 2009; see text for definitions of DU and DT. Nonadditivity
occurs whereDU 6¼ 0 orDT 6¼ 0. Contour intervals are 10 m s21 and
10 K, with blue (red) shading for negative (positive) values ex-
ceeding 5 m s21 and 5 K in magnitude.
1 While it is possible that in situ generation of planetary waves
could be occurring in the uppermesosphere as a result of baroclinic
instability, for example, the sign of the resolved wave drag gener-
ated by such waves would be opposite in sign (i.e., positive) to that
found in the simulation without OGWD. Consequently, the dom-
inant contribution to the resolved wave drag in the upper meso-
sphere in this simulation appears to be from planetary waves
propagating up from below.
JULY 2013 MCLANDRES S ET AL . 2165
understand why this happens, we resort again to down-
ward control to diagnose the contributions from the dif-
ferent types of wave drag. Since the nonadditivity is
strongest prior to the major warming (Fig. 11), we focus
on that period and time average the results to further
simplify the analysis. Figure 12 shows the GW additivity
anomaly associated with the polar-cap residual vertical
velocity Dw* (thick curve) averaged from 1 November to
31 January. Above about 55 km Dw* is negative, with a
peak value at about 65 km. The negative Dw* causes adi-abatic warming, thus explaining the positive DT in Fig. 11.
The causes of nonzero Dw* can be understood by
constructing the DC estimate of Dw* from (1); that is,
Dw5DwGWD2 (DwOGWD1DwNGWD), (2)
where w 5 w(E) 1 w(N) 1 w(O), with the bracketed su-
perscripts E, N, and O denoting the respective contri-
butions from resolved wave drag (denoted EPFD for
Eliassen–Palm flux divergence), OGWD, and NGWD
to the DC estimate ofw*. Using (2) and the equation for
w yields the DC contributions to Dw from the three
types of wave drag, which are given by
Dw(E) 5Dw(E)GWD2 (Dw(E)
OGWD 1Dw(E)NGWD), (3)
Dw(O) 52Dw(O)NGWD , (4)
Dw(N) 52Dw(N)OGWD . (5)
The simpler forms of (4) and (5), as compared to (3),
follow from the fact that two of the terms in (2) cancel
because the DC estimate of the OGWD or NGWD con-
tribution to w is zero in the simulations without OGWD
or NGWD and without both types of GWD.
The results of the decompositions (3)–(5) are given by
the thin curves in Fig. 12, where the dotted, dashed, and
dashed–dotted curves denote, respectively, the EPFD,
OGWD, and NGWD contributions to w. To demon-
strate the validity of this decomposition, the DC esti-
mate Dw is also plotted (thin solid curve), and is seen to
be in good qualitative agreement with Dw* (thick solid
curve). Above about 70 km the dominant contributions
are from NGWD and EPFD, with the two being of op-
posite sign, which explains the diminishing magnitude of
Dw* with height in that region. The maximum negative
values of Dw* near 65 km arise from the nearly equal
and negative contributions from all three wave drag
components, while the positive values of Dw* near 55 km
are primarily from EPFD.
The physical mechanisms responsible for the con-
tributions of the three types of wave drag to the non-
additivity documented in Fig. 12 can be understood by
examining the changes in wave drag between the differ-
ent simulations. Here we discuss only NGWD and EPFD
since overall they provide the dominant contributions to
Dw in the upper mesosphere shown in Fig. 12. From (5),
the NGWD contribution to Dw is simply the negative of
the response of w(N) to OGWD (i.e., w(N) from the sim-
ulation without OGWD minus w(N) from the control
simulation). Figure 10b shows w(N) for those two simu-
lations (thin solid blue and red curves) for a slightly dif-
ferent time period than used in Fig. 12. The difference
between those two curves is negative, which yields the
negative contribution fromNGWD toDw seen in Fig. 12.
The physical mechanism for this change in NGWD was
discussed in the previous section—namely, the increase
in westward NGWD in the upper mesosphere that oc-
curs in response to the stronger westerlies in the meso-
sphere when OGWD is turned off.
The EPFD contribution to nonadditivity in Fig. 12 is
more complicated since, as can be seen from (3), it must
be derived from all four simulations. Consequently, it
will not be analyzed in detail as for NGWD. The physical
mechanism responsible for its contribution can be un-
derstood simply as resulting from the strong dependence
of resolved wave drag on the zonal-mean winds and
temperatures, which in turn depend strongly upon the
OGWD and NGWD.
The fact that the zonal wind and temperature responses
to OGWD and NGWD are nonadditive is not entirely
surprising given that both the parameterized and resolved
wave drag are strongly dependent upon the background
FIG. 12. GW additivity anomaly for polar-cap residual vertical
velocityDw* (thick solid) averaged from 1Nov 2008 to 31 Jan 2009,
Dw (thin solid), and the decomposition of Dw into its components:
resolved wave drag (EPFD; dots), orographic GWD (dashed), and
nonorographic GWD (dashed dots). Nonadditivity occurs where
Dw* 6¼ 0 and Dw 6¼ 0 . See text for definitions of Dw* and Dw.
2166 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
winds. The nonadditive responses simply indicate that the
different types of wave drag interact strongly with each
other through their influence on the background state.
4. Summary and discussion
Aversion of the CanadianMiddleAtmosphereModel
that is nudged toward reanalysis data up to 1 hPa is used
to examine the relative roles of orographic and non-
orographic gravity wave drag in determining the zonal-
mean circulation of the mesosphere during the extended
northern winters of 2006 and 2009. These years are cho-
sen because of the occurrence of two of the largest
stratospheric sudden warmings on record, which strongly
coupled the lower and upper atmosphere through ver-
tically propagating planetary-scale Rossby waves and
small-scale gravity waves. By examining the extended
period from September to April we are able to study
not only the long recovery period following thewarmings,
which has been the focus of previous studies, but also the
fall/early-winter period leading up to the warmings when
the middle atmosphere was much closer to its climato-
logical state.
Four sets of simulations are performed. The first is
a control simulation in which both the OGWD and
NGWD parameterizations are active. The three others
are sensitivity experiments in which the GWD param-
eterizations are turned off, separately and together.
Each set of simulations consists of three members ini-
tialized on 1 January 2004 using perturbed initial con-
ditions. The use of ensembles of three enables us to
assess the degree to which the mesospheric circulation is
slaved to the state of the lower atmosphere.
We validate our simulations using satellite observa-
tions of temperature, zonal wind (derived from geo-
potential height), and CO from the Aura MLS
instrument. The control simulation is shown to be in
remarkably good agreement with MLS, with RMS
temperature and wind differences of less than 5 K and
5 m s21 above the top of the nudging region between
50 and 85 km. The simulated CO is also shown to be in
good (to within 10%–20% at 60 km) agreement with
MLS, and exhibits the same temporal variation over
the course of the two extended winters—namely, a two-
peaked structure, with a small late-fall/early-winter
peak and a larger late-winter peak a month or so after
the major warmings.
The sensitivity experiments reveal that the relative
roles of OGWD and NGWD are very different between
fall/early winter and late winter. In the months leading
up to the warmings, OGWD is shown to have a larger
overall impact than NGWD.During this time period the
simulation without OGWD exhibits excessively strong
(with respect to MLS) zonal-mean zonal winds, a large
amplitude temperature difference dipole in the vertical,
and overly large values of CO above about 60 km.
NGWD, on the other hand, is shown to have a more
pronounced effect in the month after the warmings as
the vortex is recovering, which is in agreement with the
findings of Ren et al. (2011). In the simulation without
NGWD the descent of the elevated stratopause is de-
layed until the lower-level zonal winds have increased to
the point where the parameterized orographic GWs are
able to reach the mesosphere, in agreement with the
findings of Hitchcock and Shepherd (2013). The pres-
ence of OGWD in late winter (a month or so after the
warmings) also drives the realistic descent of CO at that
time. In the simulation without NGWD there is much
weaker descent following the warming and conse-
quently a much weaker late-winter peak of CO in the
lower mesosphere. In the simulation without any GWD,
the late-fall/early-winter winds in the mesosphere are
extremely strong and the stratopause does not descend
from high altitudes after the warming but reforms later
than observed, in early spring, at its climatological height
as a result of increased solar heating.
A closer examination of the simulations reveals that
there is a strong interplay between the OGWD and
NGWD in late fall/early winter, with the strength of the
upper-mesospheric NGWD being controlled by the
strength of the upper-stratospheric/lower-mesospheric
westerlies, which are in turn controlled by the OGWD.
In the simulation without OGWD anomalously strong
downwelling occurs in the upper mesosphere, which ex-
plains the overly large values of lower-mesospheric
CO and the anomalously high temperatures in the upper
mesosphere in this simulation. This occurs because of an
increase in westward NGWD in the upper mesosphere
in response to the increased westerlies in the lower me-
sosphere brought about by the removal of OGWD.
The lack of spread in the zonal-mean temperature and
zonal wind time series between the different members
of each ensemble demonstrates just how strongly the
zonal-mean circulation of the mesosphere is slaved to
the lower atmosphere. Moreover, in the late fall/early
winter of 2006 and 2009 this occurs through a combination
of OGWD and resolved wave drag, while in late winter
following the major warmings it is through NGWD. An
important point to note is that the slaving of the meso-
sphere to the stratosphere does not imply that the meso-
spheric circulation is realistic, only that its temporal
variation is constrained. The realism of the mesospheric
circulation depends on the realism of the slaving mech-
anisms. This is demonstrated in the sensitivity experi-
ments in which the different types of GWD are turned
off—the members of each of those ensembles exhibit
JULY 2013 MCLANDRES S ET AL . 2167
little spread but are substantially different from the
observations (Fig. 2).
We also examine the additivity of the responses to
OGWD and NGWD, and find that they are strongly
nonadditive in the upper mesosphere before the major
warmings when OGWD, NGWD, and resolved wave
drag are simultaneously active. That the responses are
nonadditive is not too surprising given that the breaking
criterion used in the GWD parameterizations has a
strong dependence on the background winds. Moreover,
turning off one or both of the GWD terms alters the
zonal-mean winds, thus changing the propagation and
breaking of planetary waves. Nonadditivity also has
the practical implication that the OGWD and NGWD
schemes cannot be tuned separately.
We turn now to a brief discussion of some possible
implications of our study. The first is that the good
agreement between the control simulation and the MLS
observations over a wide range of NH winter conditions
provides a measure of confidence in the GWD param-
eterizations employed in CMAM, at least in terms of
their zonal-mean effects. (In the case of OGWD it must
be remembered that the local values can far exceed that
of the zonal mean, so our study does not address the
realism of the local values of OGWD.) The two-wave
OGWD scheme used here was shown by Scinocca and
McFarlane (2000) to transport 30%–50% more momen-
tum flux into the middle atmosphere in winter than the
simpler single-wave McFarlane (1987) scheme. Our re-
sults suggest that this increase in middle-atmosphere
OGWD in the Scinocca–McFarlane scheme is realistic
and that the use of two waves to represent the full spec-
trum of orographic GWs is an improvement over simpler
one-wave OGWD schemes.
Strictly speaking, the good agreement between the
control simulation and the MLS observations only ap-
plies to the total GWD, and there could in principle be
compensating errors between the OGWD and NGWD
schemes. However, as seen in Figs. 5 and 6, the agree-
ment holds both before the major warmings, when
OGWD is dominant, and after the major warmings,
when the weak lower-stratospheric winds filter most of
the OGWD, leaving the NGWD dominant at meso-
spheric heights. The different zonal wind conditions
expose themesosphere to different relative influences of
OGWD and NGWD, suggesting that the two GWD
components must each be fairly accurate.
A second implication of our study concerns data as-
similation. Ren et al. (2011) showed that it was not
necessary to assimilate mesospheric measurements to
reproduce the zonal-mean mesospheric temperature
response to the 2006 major SSW and the prolonged de-
scent of the elevated stratopause following the warming.
Our study supports that finding, and extends it to include
the entire period from September to April for the win-
ters of 2006 and 2009. The benefit of assimilating meso-
spheric data would therefore seem to lie in reproducing
zonal asymmetries in the mesosphere, which are not in-
vestigated here, or the mesospheric state during sum-
mertime,which is prone to in situ instabilities (e.g., Plumb
1983) and thus is not expected to be as strongly slaved to
the stratosphere.
A final remark is that our modeling approach could be
used to better constrain the source parameters of NGWD
parameterizations. While this suggestion has been made
before in the context of data assimilation (e.g., Ren et al.
2011), nudging to reanalysis data makes it possible to
perform a suite of simulations using a variety of different
GWD parameter settings, which would be prohibitively
expensive with a data assimilation system.
Acknowledgments. The authors thankWilliam Daffer
(JPL) for his help in preparing the MLS data and Norm
McFarlane for helpful comments on an earlier version of
the manuscript, as well as two anonymous reviewers. CM
thanks Andreas Jonsson and Diane Pendlebury for help-
ful discussions. This work was funded by the Canadian
Space Agency through the CMAM20 project.
REFERENCES
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis:
Configuration and performance of the data assimilation sys-
tem. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/
qj.828.
de Grandpr�e, J., S. R. Beagley, V. I. Fomichev, E. Griffioen, J. C.
McConnell, A. S.Medvedev, and T.G. Shepherd, 2000:Ozone
climatology using interactive chemistry: Results from the
Canadian Middle Atmosphere Model. J. Geophys. Res., 105
(D21), 26 475–26 491.
Eckermann, S. D., and Coauthors, 2009: High-altitude data assimi-
lation system experiments for the northern summer meso-
sphere season of 2007. J. Atmos. Sol.-Terr. Phys., 71, 531–551.
Haynes, P. H., C. J. Marks, M. E. McIntyre, T. G. Shepherd, and
K. P. Shine, 1991: On the ‘‘downward control’’ of extratropical
diabatic circulations by eddy-induced mean zonal forces.
J. Atmos. Sci., 48, 651–678.Hitchcock, P., and T. G. Shepherd, 2013: Zonal-mean dynamics
of extended recoveries from stratospheric sudden warmings.
J. Atmos. Sci., 70, 688–707.
——, ——, and G. L. Manney, 2013: Statistical characterization of
Arctic polar-night jet oscillation events. J. Climate, 26, 2096–
2116.
Jin, J. J., and Coauthors, 2009: Comparison of CMAM simulations
of carbon monoxide (CO), nitrous oxide (N2O), and methane
(CH4) with observations from Odin/SMR, ACE-FTS, and
Aura/MLS. Atmos. Chem. Phys., 9, 3233–3252, doi:10.5194/
acp-9-3233-2009.
Jonsson, A. I., J. de Grandpr�e, V. I. Fomichev, J. C. McConnell,
and S. R. Beagley, 2004: Doubled CO2-induced cooling in
the middle atmosphere: Photochemical analysis of the ozone
2168 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 70
radiative feedback. J. Geophys. Res., 109,D24103, doi:10.1029/
2004JD005093.
Limpasuvan, V., J. H. Richter, Y. J. Orsolini, F. Stordal, and O.-K.
Kvissel, 2012: The roles of planetary and gravity waves during
a major stratospheric sudden warming as characterized in
WACCM. J. Atmos. Sol.-Terr. Phys., 78–79, 84–98, doi:10.1016/
j.jastp.2011.03.004.
Livesey, N. J., and Coauthors, 2011: Earth Observing System (EOS)
Aura Microwave Limb Sounder (MLS) Version 3.3 Level 2
data quality and description document. Jet Propulsion Labo-
ratory Tech. Rep. JPL D-33509, 156 pp. [Available online at
http://mls.jpl.nasa.gov/data/v3-3_data_quality_document.pdf.]
Manney, G. L., and Coauthors, 2008a: The evolution of the
stratopause during the 2006 major warming: Satellite data
and assimilated meteorological analyses. J. Geophys. Res.,
113, D11115, doi:10.1029/2007JD009097.
——, and Coauthors, 2008b: The high Arctic in extreme winters:
Vortex, temperature, and MLS and ACE-FTS trace gas
evolution. Atmos. Chem. Phys., 8, 505–522, doi:10.5194/
acp-8-505-2008.
——, and Coauthors, 2009a: Satellite observations and modeling
of transport in the upper troposphere through the lower
mesosphere during the 2006 major stratospheric sudden
warming. Atmos. Chem. Phys., 9, 4775–4795, doi:10.5194/
acp-9-4775-2009.
——, and Coauthors, 2009b: Aura Microwave Limb Sounder ob-
servations of dynamics and transport during the record-
breaking 2009 Arctic stratospheric major warming. Geophys.
Res. Lett., 36, L12815, doi:10.1029/2009GL038586.
McFarlane, N. A., 1987: The effect of orographically excited
gravity wave drag on the circulation of the lower stratosphere
and troposphere. J. Atmos. Sci., 44, 1775–1800.
Merryfield, W. J., and Coauthors, 2013: The Canadian Seasonal to
Interannual Prediction System. Part I: Models and initializa-
tion. Mon. Wea. Rev., in press.
Plumb, R. A., 1983: Baroclinic instability of the summer meso-
sphere: a mechanism for the quasi-two-day wave? J. Atmos.
Sci., 40, 262–270.
Polavarapu, S., S. Ren, Y. Rochon, D. Sankey, N. Ek, J. Koshyk,
and D. Tarasick, 2005: Data assimilation with the Canadian
Middle Atmosphere Model. Atmos.–Ocean, 43, 77–100.
Randall, C. E., V. L. Harvey, C. S. Singleton, P. F. Bernath, C. D.
Boone, and J. U. Kozyra, 2006: Enhanced NOx in 2006 linked
to strong upper stratospheric Arctic vortex. Geophys. Res.
Lett., 33, L18811, doi:10.1029/2006GL027160.
——,——,D. E. Siskind, J. France, P. F. Bernath, C.D. Boone, and
K. A. Walker, 2009: NOx descent in the Arctic middle atmo-
sphere in early 2009.Geophys. Res. Lett., 36,L18811, doi:10.1029/
2009GL039706.
Randel, W. J., 1987: The evaluation of winds from geopotential
height data in the stratosphere. J. Atmos. Sci., 44, 3097–
3120.
Ren, S., S. Polavarapu, and T. G. Shepherd, 2008: Vertical propa-
gation of information in a middle atmosphere data assimila-
tion system by gravity-wave drag feedbacks. Geophys. Res.
Lett., 35, L06804, doi:10.1029/2007GL032699.
——, ——, S. R. Beagley, Y. Nezlin, and Y. J. Rochon, 2011: The
impact of gravity wave drag on mesospheric analyses of the
2006 stratospheric major warming. J. Geophys. Res., 116,
D19116, doi:10.1029/2011JD015943.
Scinocca, J. F., 2003: An accurate spectral nonorographic gravity
wave drag parameterization for general circulation models.
J. Atmos. Sci., 60, 667–682.
——, and N. A. McFarlane, 2000: The parametrization of drag
induced by stratified flow over anisotropic orography. Quart.
J. Roy. Meteor. Soc., 126, 2353–2393.
——, ——, M. Lazare, J. Li, and D. Plummer, 2008: The CCCma
third generation AGCM and its extension into the middle
atmosphere. Atmos. Chem. Phys., 8, 7055–7074, doi:10.5194/
acp-8-7055-2008.
Shepherd, T. G., and T. A. Shaw, 2004: The angular momentum
constraint on climate sensitivity and downward influence in
the middle atmosphere. J. Atmos. Sci., 61, 2899–2908.
Siskind, D. E., S. D. Eckermann, J. P. McCormack, L. Coy, K. W.
Hoppel, and N. L. Baker, 2010: Case studies of the mesospheric
response to recent minor, major, and extended stratospheric
warmings. J. Geophys. Res., 115, D00N03, doi:10.1029/
2010JD014114.
SPARC CCMVal, 2010: SPARC Report on the evaluation of
chemistry-climate models. SPARC Rep. 5, WCRP-132,
WMO/TD-No. 1526, 434 pp. [Available online at http://www.
atmosp.physics.utoronto.ca/SPARC/ccmval_final/index.php.]
JULY 2013 MCLANDRES S ET AL . 2169